Bhanu Prakash
Location: Hyderabad, India
Phone: (+91) 9182841695
Email: ml.bhanuprakash@gmail.com
LinkedIn: linkedin.com/in/ml-bhanuprakash
GitHub: github.com/ml-bhanuprakash
📄 [Download Resume (PDF)]
🧠 Profile Summary
AI Engineer with over 3 years of industrial and collaborative experience in developing advanced AI solutions.
Led multiple simulator-based projects involving environment creation, texture design, self-driving agents, and Sim2Real pipelines, collaborating closely with simulator teams to bridge technical gaps and deliver robust results.
Experienced in autonomous vehicle systems, including data acquisition, live testing, and performance evaluation of self-driving tractors using ROS and stereo camera setups.
Proficient in Deep Learning, Transformers, Vision-Language Models, PyTorch, and OpenCV, with a focus on end-to-end AI systems for computer vision and multimodal applications.
⚙️ Skills
Core Areas: Deep Learning, Computer Vision, Generative AI, Transformers, Vision-Language Models
Tech Stack: YOLO, Image Segmentation, SfM, 3D Reconstruction, ROS, Sim2Real, Stereo Camera, PyTorch, OpenCV, LangChain, LangGraph, Python
🚀 AI Projects
🔗 Website | Docs
Built an AI-powered web platform that enables text + image-based outfit discovery and custom designer booking for Indian traditional fashion. The system integrates the SigLIP Vision Transformer for cross-modal embeddings and FAISS for high-speed similarity search.
A Flask backend handles data automation and designer assignment workflows, supported by a structured JSON dataset that links embeddings, FAISS indices, and product records via unique IDs.
Enhanced query relevance using Ollama (Mistral LLM) for SEO-optimized text generation and deployed the entire pipeline with Docker on Hugging Face Spaces, ensuring scalability and secure dataset management.
Demonstrates expertise in Vision Transformers, LLM pipelines, vector search, and full-stack AI deployment.
CCTV to Stereo – 3D Vision System Built from Hikvision Cameras
🔗 Documentation
Built a custom stereo vision system using two Hikvision 2MP IR Bullet Cameras for indoor calibration and outdoor vehicle testing. Performed stereo calibration, rectification, and depth estimation using OpenCV and addressed real-world challenges like power stability, voltage variation, and live monitoring via a TP-Link router with OpenWRT.
Mounted on a vehicle, the system validated dynamic depth estimation and pixel correspondences, providing hands-on experience in stereo geometry, real-time depth mapping, embedded networking, and automotive deployment. The project includes detailed documentation and demo videos for reference.
End-to-End YOLO Object Detection Pipeline (Data to Android Deployment)
🔗 Demo
Built a custom YOLO annotation tool with features to draw, add, and delete object classes. Added image enhancement for low-light/night images for accurate labeling.
Converted the model to TFLite and deployed a real-time Android object detection app.
Self-Driving Car with Lane Detection using Arduino
🔗 Documentation | Demo
Implemented real-time lane detection using Pixel Summation, thresholding, and perspective warping to calculate road curvature. Integrated with Arduino, TB6600 stepper driver, and NEMA 17 motor to automate steering, demonstrating hands-on experience in computer vision, robotics, and embedded systems.
💼 Work Experience
3D Wedding Reconstruction & VR Visualization
📅 Dec 2024 – Present
🔗 Documentation
- Led end-to-end technical execution of a research project combining computer vision, 3D reconstruction, and VR visualization to create interactive wedding experiences inspired by GTA-style open-world navigation.
- Collaborated with a professional Telugu film editor for creative direction, ensuring high-quality cinematic storytelling using original video footage.
- Designed and managed multi-camera setups (DSLRs, drones, smartphones, CCTV rigs) to capture multi-angle, high-overlap footage optimized for cost and scalability.
- Built dynamic and static 3D point clouds using COLMAP, Agisoft Metashape, RealityCapture, Meshroom, Pix4Dmapper, and OpenMVG/OpenMVS; optimized camera poses and processing outputs for VR-ready visualization.
- Developed multi-camera rigs for synchronized moving point clouds, capturing human gestures and interactions while exploring real-time motion reconstruction limits.
- Converted reconstructed scenes into interactive VR/HTML environments, applied cinematic stylization with Adobe After Effects, and generated print-ready 3D models for demonstrations.
Perception Engineer – Monarch Tractor (Zimeno India Private Limited)
📅 April 2023 – Nov 2024
Tools & Technologies: Python, PyTorch, OpenCV, Unreal Engine, Stereo Vision, Object Detection, Segmentation, Depth Estimation, Point Clouds, EXE Tool Development
Key Responsibilities & Achievements:
- Led AI perception development for autonomous tractors, bridging AI and Simulator teams to align real-world and virtual data workflows; includes demo video of the autonomous agent driving itself in simulation (Demo Link).
- Developed and trained object detection, segmentation, and stereo vision models to enhance tractor perception in complex agricultural environments.
- Built Camera Clarity models for real tractors to detect mud, blur, water droplets, and glare, ensuring reliable perception under challenging conditions.
- Planned and designed Sim2Real adaptation, converting simulator (game-engine-based) visuals into realistic data for model testing across diverse environments and perception tasks.
- Developed custom EXE-based annotation tools for Object Detection, Segmentation, and Point Cloud labeling. Managed annotation teams (in-house and outsourced), handling task assignment, progress tracking, and quality assurance.
- Conducted tractor stoppage diagnostics, identifying root causes of unexpected stops — including perception errors, autonomous decision faults, control issues, or software failures — and implemented solutions to reduce false detections.
- Processed and aligned 3D point clouds from real-world and simulated sources; generated 3D meshes in the simulator to recreate realistic field environments for AI model training and validation.
- Built and maintained a Data Management Tool to automate dataset ingestion, cleaning, deduplication, and metadata tagging — improving data accessibility and storage efficiency across teams.
Tailortech Private Limited (CloudTailor), Hyderabad – Machine Learning Engineer
📅 April 2022 – Mar 2023
Online Human Body Measurements
🔗 Demo
- Solely designed and implemented a complete end-to-end computer vision system to automatically extract human body measurements from front and side images.
- Used MediaPipe to detect precise body landmarks and compute measurements (shoulder, bust, arm, thigh, waist, hip) along with height, assuming proper standing posture.
- Enabled fully automated measurement extraction when the person stands correctly, reducing manual intervention and improving accuracy.
- Delivered a personalized styling service that recommends outfits based on accurate measurements and reference images, achieving 98% measurement accuracy.
Cloth Pattern Detector
- Developed an image recognition and reverse search tool using ResNet-50 embeddings to identify clothing and accessories from reference images and rank results by similarity.
- Extended search to the web for additional relevant items, enabling personalized fashion recommendations for rapidly changing trends.
📝 Declaration
I hereby declare that the details and information given above are complete and true to the best of my knowledge.
Name: B. Bhanu Prakash
Place: Hyderabad, India